A style LoRA for **Anima** reproducing a bold, high-contrast doujin-illustration look: heavy expressive lineart, exaggerated facial proportions, strong tonal contrast and confident compositions. ## How to use ``` chaozhi, 1girl, ``` **Best-known combination:** `chaozhi` 0.8 + `koushui` 0.6 — in our A/B tests this pairing gave the largest simultaneous gain in edge definition **and** saturation (Δedge +1.14, Δsat +6.54 versus base). ## Recommended settings | Item | Value | |-|-| | Base model | Anima (aesthetic v1.x) | | LoRA weight | 0.8 (start here; 1.0 for maximum effect) | | Steps | 25–35 | | CFG | 5–7 | ## Known limitations - Above 1.0 faces start to distort; 0.8 is the sweet spot for anatomy. - It is a *style* LoRA — it carries no characters. Combine with a character LoRA for specific subjects. - Trained on colour illustrations only; it cannot render convincing monochrome manga pages. ## Training details 63 curated colour illustrations · 1024px · Anima base. v1.1 = a single low-LR (2e-6) fine-tune epoch from the v1.0 convergence point. Style LoRA trained for personal use. If you are the original artist and want this removed, contact me and I will take it down.